Fabian Zills

https://github.com/zincware/ZnTrack/blob/e0f610739d2e05c0093f685fb050b7df62e5b5cf/zntrack/project/zntrack_project.py#L222 ```python with project.create_experiment(): new_node = MyNode() ``` should be possible by entering `self.graph`

If one type of Node occurs multiple times, consider combining the plots automatically: ```yaml plots: - nodes/MLModel/training.csv: template: dvc_plots/templates/y_log.json x: epoch y: - loss - rmse_e_per_atom - rmse_f - nodes/MLModel_1/training.csv:...

Especially with `automatic_node_names` it would be nice to have an easy way to rename nodes.

Something like: `data = {25: A, 13: B}` would be nice. This would require to split `zn.deps` and create ```yaml data: - 25 - 15 ``` which could be related...

Convert custom file formats

1

use some parallel call, or during the subprocess call to convert metrics files which arent dvc compatible to DVC compatible ones.

There should be a `_type` key for every entry in `connections`: ```json "_type": "znflow.CombinedConnections", "value": { "connections": [ { "instance": { "_type": "zntrack.Node", "value": { "module": "ipsuite.data_loading.add_data_ase", "cls": "AddData", "name":...

zn.deps order

1

The following two should be different: ```python Node(data=[data1, data2]) != Node(data=[data1, data2]) ```

Maybe add `dataclasses_for_parameters.ipynb` a version where you use a Node instead.

I find myself writing `outs: zn.outs()` occasionaly and it is difficult to debug. It would be nice to log a warning if this happens.